Modeling software failures and reliability growth during system testing
Willa K. Ehrlich, TercassA Emerson · International Conference on Software Engineering · 1987
A number of time-domain software reliability models attempt to predict the growth of a system's reliability during the system test phase of the development life cycle. In this paper we examine the results of applying several types of Poisson-process models to the development of a large system for which system test was performed in two parallel tracks, using different strategies for test data selection. We show that the reliability growth predicted by non-homogeneous Poisson process models was found for only one of these testing strategies. These results imply that the applicability of a reliability growth model to a given software development project will depend on the nature of that project's system test process; they also raise theoretical questions about the assumption of certain statistical properties for failure occurrence during testing.